AI could prevent hundreds of construction deaths in Bangladesh, study suggests

A recent incident in Bandarban, Bangladesh, where two construction workers were electrocuted while handling a metal bar that contacted a live power line, underscores a persistent and deadly problem in the country's construction sector. Sakib, 20, and Kamrul Hasan, 35, became the latest casualties in a nation where workplace accidents claim the lives of well over a hundred construction workers annually, according to safety researchers who deem many of these tragedies preventable.

These are not isolated events. A survey by the Safety and Rights Society (SRS) found 802 worker deaths from workplace accidents across Bangladesh in 2025, with 120 in construction alone. The situation worsened in early 2026, as the Human Rights Support Society (HRSS) reported 72 worker fatalities in the first three months, nearly triple the number from the same period in 2025. A decade-long review by researchers from Concordia University and Rajshahi University of Engineering and Technology documented over 1,400 construction-related deaths, identifying falls from height and electrocution as the primary causes.

Bangladesh has established safety regulations, including the National Building Code and the Labour Act, which mandate protective gear like helmets and harnesses, and require properly built scaffolding. However, enforcement remains a challenge, as manual inspections cannot constantly monitor every worker on every site. Site engineers, even diligent ones, cannot be everywhere at once, making inspections inherently reactive rather than proactive.

The article proposes that artificial intelligence and computer vision technology could bridge this safety gap. Cameras already used for security could be integrated with deep learning models to continuously monitor video feeds. These systems could detect when workers are not wearing required safety equipment or are in proximity to dangerous machinery or unprotected edges, issuing alerts within seconds. Systems using object-detection models like YOLO have shown high accuracy rates, up to 96 percent for identifying missing protective gear, and fall-prevention systems report precision close to 99 percent. While AI-powered safety tools are being adopted by large construction firms in North America and Europe, and AI cameras are being used for retail and traffic surveillance in Bangladesh, their deliberate application to the country's most hazardous workplaces is notably absent.